search

chenyuntc / simple-faster-rcnn-pytorch

A simplified implemention of Faster R-CNN that replicate performance from origin paper
Other
3.99k stars 1.14k forks source link

Loss leaps after every epoch #142

Open YangYangGirl opened 5 years ago

YangYangGirl commented 5 years ago

image

YangYangGirl commented 5 years ago

I wrote the poink-linking-network based on your code,but the loss looked strange which leaped after every epoch.So I had took a close look out at your loss,I found that there also was a relatively weak leaping after every epoch.I felt very strange about it. https://github.com/YangYangGirl/point-linking-network

300LiterPropofol commented 4 years ago

The leap is normal because the current repository only supports batch_size = 1 training, it means it randomly select out one image instead of a mini batch that includes multiple images, this randomness causes the leap after every epoch, if you change the batch_size to 8 our 16, the leap will disappear.